【摘要】 针对木制家具的表面死节缺陷,提出一种基于正则化共面判别分析(RCDA,Regularized Coplanar Discriminant Analysis)与支持向量机(SVM,Support Vector Machine)的死节缺陷图像分割算法。将RGB彩色图像转换成灰度图像,对灰度图像进行分块,同时将块变换成列向量,所有列向量组成矩阵进行RCDA维数约减,对约减后的特征进行SVM训练与测试,得到图像块分类结果。最后将块分类矩阵变形成二值分割图,得到死节缺陷目标。试验结果表明,提出的算法效果好,SD、Dice、ER、NR值分别为80. 96%、89. 48%、23. 33%、0. 16%。更多还原

【Abstract】 Aiming at the dead joint defects on wooden furniture surface,a dead knot image segmentation algorithm based on regularized coplanar discriminant analysis( RCDA) and support vector machine( SVM) was proposed. An RGB color image is transformed into a gray image,and the gray image is divided into blocks. At the same time,the blocks are transformed into column vectors. All column vectors form a matrix for RCDA dimension reduction,and SVM training and testing are performed on the reduced features to obtain image block classification results. Finally,the block classification matrix is transformed into a binary segmentation graph to obtain the dead knot defect target.The experimental results show that the proposed algorithm works well,with the SD,Dice,ER and NR values being80. 96%,89. 48%,23. 33% and 0. 16%,respectively.更多还原